Nonlinear multivariate time series (NLMVTS) analysis of whole tree fibre dimensions in E-globulus

Prediction equations confirm a significant and predominant Oblique I directionality following attained height (AH) lines from pith to bark and up the stem in I E, globulus tree (G3). The differential change in fibre morphology (with cambial age (CA(1)) or ring (YR)) along these Oblique I axes is consistent between AIIs (YR*AH, P > 0.2). Within-ring changes in fibre length (FL), fibre coarseness (CR) and fibre diameter follow varying CA up the stem in G3; but this pattern of change varies significantly between rings (CA*YR, 0.005 < P < 0.05). The within-ring longitudinal variation (Oblique axis 11) of FL, CR, fibre diameter, lumen diameter and wall thickness, when referenced to height (HT) or AH, is consistent between years in G3 (YR* HT, P > 0.5). Within-ring variation in FL, shows a significantly flatter decline when referenced to AH compared to HT. Seasonal earlywood (EW) to latewood (LW) growth affects within-ring variation of FL (P < 0.0005); with this EW to LW variation consistent between rings (and heights). FL is the only fibre dimension where this seasonal effect is a significant predictor of whole tree variation. Variation up the stem, at a given number of rings from the pith, shows predominantly a significant, non-cylindrical symmetry; whose pattern changes as the number of rings from the pith (CA) changes (CA*HT, CA*AH; 0.005 <P < 0.05). On a whole tree basis, radial variation of all 5 fibre dimensions is consistent between heights. However, when radial variation is referenced to the AH of the rings traversed rather than year (YR), or when Oblique I variation along a given AH is referenced with respect to (w.r.t.) HT, there is a significant AH*HT interaction (0.0005 < P < 0.05), indicative of a steeper change, radially and obliquely, in all fibre dimensions (except lumen diameter) in the lower stem. Within-ring height partitions, which are translatable to whole tree height partitions, have been confirmed in G3 by a time series prediction approach.